towards analysing non-determinismin bidirectional transformations (amt2014)

Romina Eramo Romeo Marinelli, Alfonso Pierantonio, Gianni Rosa

Towards Analysing Non-‐Determinism

in Bidirec?onal Transforma?ons

Dipartimento di Ingegneria e Scienze

Università degli Studi dell’Aquila

dell’Informazione e Matematica

Goal

To support designers in solving non-‐determinism in their specifica?ons by detec?ng the rules that give place to alterna?ve solu?ons, at design ?me.

Towards Analysing Non-Determinism in Bidirectional Transformations – Romina Eramo

Roadmap

Background

Contribu?on

Realiza?on

Applica?on

Challenges

Conclusion


Bidirec?onality

Bidirec?onal transforma?ons are useful for maintaining the consistency of two (or more) related sources of informa?on The relevance of bidirec?onality in model transforma?ons has been advocated already in 2005 by OMG by including a bidirec?onal language in QVT


Bidirec?onality

In spite of its relevance, bidirec?onality has rarely produced an?cipated benefits

WHY?


Non-‐determinism

Probably, among the reasons why bidirec?onal transforma?on had limited success there is the ambivalence concerning non-‐bijec)vity When transforma?on are non-‐bijec)ve, there may be mul?ple ways to transform two models into a consistent state


A small example ...


...a typical round-‐trip problem based on a non-‐bijec?ve class diagram to rela?onal data base -‐ UML2RDBMS

A small example ...


Person

name {is_primary}

Worker

working_address

Employer {persistent}

department

number_of_publication

Professor {persistent}

University

① 

UML model

A small example ...


Person

name {is_primary}

Worker

working_address


department



University

PK name working_address

Employer

department


University

Professor PK name

working_address department

①  ②  forward transf.

UML model

RDBMS model

Persistent classes are mapped to correspondent tables and a\ributes to columns, including inherited a\ributes A\ributes of non-‐persistent classes are distributed over tables from persistent classes which access non-‐persistent ones

A small example ...


Person

name {is_primary}

Worker

working_address


department



University


Employer

department


University

Professor PK name


①  ② 


Employer

department


University

Professor PK name


③ 

email

forward transf. manual change

UML model

RDBMS model modified RDBMS model

A new column email is added to the table Employer How back propagate ?

University University

A small example ...


Person

name {is_primary}

Worker

working_address


department



University


Employer

department


University

Professor PK name


①  ② 


Employer

department


University

Professor PK name


③ 

Person

name {is_primary}

Worker

working_address


department



University

④ 

[email]

[email]

[email]

email

forward transf. backward transf. manual change

UML model

RDBMS model modified RDBMS model

corresponding UML models

email can be associated with Employer or with each parent Worker and Person

Related issues

There have been several works analyzing seman?c issues of bidirec?onal model transforma?ons From Stevens 2009 : “The developer needs full control of what the transforma)on does. [...] We claim that determinism is necessary in order to ensure, first, that developers will find tool behavior predictable, and second, that organisa)ons will not be unacceptably “locked in” to the tool they first use.”

[P. Stevens. Bidirec?onal model transforma?ons in QVT: seman?c issues and open ques?ons. SOSYM, 2009]


How it is solved in the current approaches?

Drawbacks:

①  limited control over how models are synchronized ②  early decision and difficult to detect ambiguity ③  onerous management of (large) set of models


①  fix ambiguity via defaults strategy (tool) ②  resolve ambiguity by contraining the specifica?on (designer) ③  genera?ng mul?ple solu?ons (tool) and select the desired one

(desinger)

Roadmap

Background

Contribu?on

Realiza?on

Applica?on

Challenges

Conclusion


Contribu?on

GOAL

HOW

OUTPUT


ADVANTAGE

To support designers in dealing with non-‐determinism

Sta?c analysis of bidirec?onal transforma?on

Degree of non-‐determinism of the transforma?on Non-‐determinis?c por?ons of the transforma?on

Analysis at design ?me Detec?on of code ambiguity before the execu?on

Roadmap

Background

Contribu?on

Realiza?on

Applica?on

Challenges

Conclusion

Descrip?on of the approach

Model transforma?ons are mapped in a format suitable for the analysis. Only relevant informa?on are captured. The analysis is performed within a logic environment realized in Answer Set Programming (ASP) •  the engine is able to analyze the behavior of the transforma?on

with an emphasis on non-‐determinism


Analysis nota?on

Logical rules and constraints

Descrip?on of the approach

STEP 1 -‐ Transla?on from the transforma?on to the analysis nota?on STEP 2 -‐ Execu?on of the ASP-‐based analysis and annota?on of the

transforma?on


BIDIRECTIONAL TRANSFORMATION

LANGUAGE

C2 C2 TRASLATE

ASP TRANSF. ANALYSIS MODEL

ASP SOLVER

IN OUT

ANNOTATE

BIDIRECTIONAL TRANSFORMATION

MODEL

TRANSFORMATION ANALYSIS LANGUAGE

TRANSFORMATION ANALYSIS MODEL

ASP ANALYSIS RESULT MODEL

ENCODE

STEP 1 STEP 2

LOGICAL ANALYSIS

ASP RULES AND CONSTRAINTS

Roadmap

Background

Contribu?on

Realiza?on

Applica?on

Challenges

Conclusion

Applica?on of the approach

UML to RDBMS, inheritance hierarchy of classes to corresponding tables, and viceversa


Case study

①  implementa?on of the UML2RDBMS bidirec?onal transf. in JTL ②  transforma?on of the UML2RDBMS transf. in the analysis language ③  execu?on of the analysis on the UML2RDBMS analysis model ④  annota?on of the UML2RDBMS transf. model with the analysis results

Used transforma?on language JTL (Janus Transforma?on Language), a constraint-‐based bidirec?onal language, able to deal with non-‐bijec?vity by genera?ng all the admissible solu?ons 4 STEPS

①  Implemen?ng UML2RDBMS in JTL


1transformation UML2RDBMS(uml:UML, rdbms:RDBMS) { ...! 2 top relation Class2Table {! 3 cn, an: String;! 4 enforce domain uml c:Class {! 5 is_persistent = true,! 6 name = cn,! 7 attrs = attr: Attribute { name = an}! 8 };! 9 enforce domain rdbms t:Table {!10 name = cn,!11 cols = col: Column { name = an }!12 };!13 when { ... }!14 where { Attribute2Column(c, t); }!15 }!16 relation Attribute2Column {!17 an, at : String;!18 enforce domain uml c:Class {!19 attrs = attr: Attribute { name = an, owner = c, is_primary = false }!20 };!21 enforce domain rdbms t:Table {!22 cols = col: Column { name = an, owner = t }!23 };!24 when { ... }!25 }!26 top relation SuperAttributeToColumn{!27 enforce domain uml c: Class {!28 parent = sc: Class { }!29 };!30 enforce domain rdbms t: Table { };!31 when { ClassToTable(c,t) | (cc = c.parentOf & SuperAttributeToColumn(cc,t));}!32 where { AttributeToColumn(sc, t); }!33 } ...!

relates classes and tables in the two different metamodels

relates albutes and columns in the two different metamodels

relates a\ributes of the parent class to columns of the corresponding child table

②  Transla?ng in the analysis nota?on


A transforma?on model consists of a set of rules that define mapping among elements types, and pre/post con??ons that limit its applicability and behavior

The correspondent transforma?on analysis model captures only the informa?on which is relevant for performing the analysis, conforming a specific analysis nota?on

domain pa\erns are captured

pre and post condi?on are captured as pa\erns

③  Execu?ng the analysis


1 are_equal_domains(ID1, ID2, Dom) :-! 2 have_equal_patterns(ID1, ID2, Dom, MC, MCName),! 3 have_equal_predicates(ID1, ID2, Dom, MC, MCName),! 4 not have_different_patterns(ID1, ID2, Dom).! 5! 6 non_deterministic_relation(ID, RelName, DomLeft) :-! 7 are_equal_domains(ID, ID2, DomRight),! 8 relation_name(ID, RelName),! 9 tranformation_model(DomRight), !10 tranformation_model(DomLeft), !11 DomRight != DomLeft.!12!13 are_ambiguous_relations(ID1, ID2, DomLeft) :-!14  are_equal_domains(ID1, ID2, DomRight),!15  tranformation_model(DomRight), !16  tranformation_model(DomLeft), !17  DomRight != DomLef. [...]!

1 relation_name(1, class2table).!2 relation_domain(1, uml).!3 relation_pattern(1, uml, c, class).!4 relation_predicate(1, uml, c, is_persistent, true).!5 relation_predicate(1, uml, c, name, cn).!6 relation_predicate(1, uml, c, attrs, attr). [...]!

the analysis model is automa?cally encoded in ASP

non-‐determinism is verified by executed ASP rules and constraints on the analysis model used

as base knowledge

%%%% sets of non-deterministic relations (RDBMS to UML direction):![(1,class2table), (3,superAttribute2column), (5,superAttribute2column)]!

sets of ambiguous transforma?on rules

are detected


④  Annota?ng the UML2RDBMS transf.


A set of non-‐determinis?c rela?ons in UML to RDBMS direc?on

is detected

Roadmap

Background

Contribu?on

Realiza?on

Applica?on

Challenges

Conclusion

Challenges


Despite the analysis is completely independent from JTL, in order to make the approach completely language-‐independent addi?onal implementa?on efforts are required

Language-‐independent approach

Analysis of formal proper?es The logical founda?on of the engine makes possible the verifica?on of different formal proper?es by transla?ng them in ASP, with the scope to provide a mean for improving the quality of model transforma?on specifica?ons

Roadmap

Background

Contribu?on

Realiza?on

Applica?on

Challenges

Conclusion

Conclusion


Bidirec?onal transforma?ons are known as a promising approach for maintaining models in a consistent state, their success is limites by seman?c issues and intrinsic characteris?cs This work represent a first step to support transforma?on implementors with a tool capable of analyzing transforma?ons in order to return feedback which could resolve poten?al problem, such as non-‐determinism

Thanks

Dipartimento di Ingegneria e Scienze

Università degli Studi dell’Aquila dell’Informazione e Matematica

Romina Eramo [email protected]

towards analysing non-determinismin bidirectional transformations (amt2014)

Documents